"Syntactic (i.e., grammatical) complexity refers to the range and the degree of sophistication of the forms that appear in language production (Ortega, 2003). This concept has long been regarded as an important construct of language proficiency and has been actively investigated in the field of second language (L2) writing. Syntactic complexity is multidimensional in nature, and there are a variety of measures that tap into different dimensions of the construct. Widely used measures of complexity (e.g., mean length of T-unit and the number of clauses per T-unit) capture a relative degree of sophistication, but do not provide measures of participants' command of a range of diverse syntactic structures. In contrast, in L2 assessment, grammatical knowledge is often evaluated in terms of both syntactic elaboration and structural variety (Rimmer, 2006). To address the gap, the present study proposes new measures that tap into the diversity dimension of syntactic complexity: types and type/token frequency of verb-argument constructions (VACs). The present study investigates whether the proposed diversity measures of syntactic complexity, in combination with currently used measures of elaboration, accurately predict L2 written proficiency and writing quality. Specific research questions that guide the study are as follows: (1) Does the syntactic complexity of Korean EFL learners' writing production, as measured by various quantitative complexity measures, function as an indicator of proficiency? In addition, does adding diversity measures increase the predictive power of syntactic complexity in discriminating proficiency levels? (2) How do different syntactic complexity measures relate to subjective ratings of writing quality judged by human raters? Which measure(s) best predict writing quality? (3) How do raters interpret the notion of syntactic complexity that appear on Language Use scale of a given analytic writing rubric? Essays were collected from 390 Korean EFL learners and analyzed using corpus analytic tools. Fourteen elaboration measures were calculated using Syntactic Complexity Analyzer, an automated computational tool developed by Lu (2010). For the diversity measures, all instances of VACs in the participants' essays were retrieved and analyzed using a part-of-speech tagging tool and a concordance tool. Thirteen VAC patterns (e.g., verb + direct object, verb + indirect object + direct object, verb + direct object + object predicative, etc.) and their sub-patterns were identified based on findings in usage-based approaches to grammar, namely, construction grammar and corpus-based descriptive grammar. Then the distribution and the number of VAC types and type/token frequency of VACs were examined. Participants' proficiency levels were independently measured by a cloze test, and the quality of their essays was evaluated by human raters. The empirical results of the study indicated that measures of syntactic complexity functioned as predictors that discriminate among different proficiency levels, and adding diversity measures of complexity increased the predictive power. The diversity measures were also found to be strong predictors of human-rated writing quality, which lend support to the use of the diversity measures in this area of research. Qualitative data obtained from the rater interviews showed that notions of grammatical complexity as interpreted by raters generally overlap with the notion of syntactic complexity in SLA. However, variability was found in the interpretations between raters."--Pages ii-iii.
Read
